If you know the initial location(s) of the animal, provide this information via .xinit (or .map). If .xinit is not provided, initial locations are sampled uniformly from the part of the .map that is compatible with the initial observations. In patter@dev, this happens in Julia. You can set a unique initial map via set_map(..., .as_Raster = TRUE, .as_GeoArray = FALSE). For custom ModelObs subtypes, map_init() methods are required for automated sampling of the part of the map compatible with observations. In patter v.1.0.0, these methods should be provided in R, in patter@dev, methods written in Julia are required.